Pharmacokinetic-Pharmacodynamic Determinants of Clinical Outcomes for Rifampin-Resistant Tuberculosis: A Multisite Prospective Cohort Study

Abstract

Background: Rifampin-resistant and/or multidrug-resistant tuberculosis (RR/MDR-TB) treatment requires multiple drugs, and outcomes remain suboptimal. Some drugs are associated with improved outcome. It is unknown whether particular pharmacokinetic-pharmacodynamic relationships predict outcome.

Methods: Adults with pulmonary RR/MDR-TB in Tanzania, Bangladesh, and the Russian Federation receiving local regimens were enrolled from June 2016 to July 2018. Serum was collected after 2, 4, and 8 weeks for each drug's area under the concentration-time curve over 24 hours (AUC0-24). Quantitative susceptibility of the M. tuberculosis isolate was measured by minimum inhibitory concentrations (MICs). Individual drug AUC0-24/MIC targets were assessed by adjusted odds ratios (ORs) for favorable treatment outcome, and hazard ratios (HRs) for time to sputum culture conversion. K-means clustering algorithm separated the cohort of the most common multidrug regimen into 4 clusters by AUC0-24/MIC exposures.

Results: Among 290 patients, 62 (21%) experienced treatment failure, including 30 deaths. Moxifloxacin AUC0-24/MIC target of 58 was associated with favorable treatment outcome (OR, 3.75; 95% confidence interval, 1.21-11.56; P = .022); levofloxacin AUC0-24/MIC of 118.3, clofazimine AUC0-24/MIC of 50.5, and pyrazinamide AUC0-24 of 379 mg × h/L were associated with faster culture conversion (HR >1.0, P < .05). Other individual drug exposures were not predictive. Clustering by AUC0-24/MIC revealed that those with the lowest multidrug exposures had the slowest culture conversion.

Conclusions: Amidst multidrug regimens for RR/MDR-TB, serum pharmacokinetics and M. tuberculosis MICs were variable, yet defined parameters to certain drugs-fluoroquinolones, pyrazinamide, clofazimine-were predictive and should be optimized to improve clinical outcome.

Clinical trials registration: NCT03559582.

Keywords: minimum inhibitory concentrations; multidrug-resistant tuberculosis; pharmacodynamics; pharmacokinetics.

Figure 1.

Change in total serum exposure. Change expressed as population mean and standard errors for the AUC_0–24 (mg × h/L) measured at 2, 4, and 8 weeks after treatment initiation as visualized for each drug at each site: Tanzania (square), Bangladesh (triangle), Russian Federation (x), and all sites averaged (circle). Total patients for each drug at 2-, 4-, and 8-week time points: pyrazinamide (266, 171, 127), kanamycin (193, 141, 88), cycloserine (163, 91, 79), levofloxacin (164, 101, 68), prothionamide (118, 69, 45), ethionamide (132, 105, 86), clofazimine (56, 13, 13), ethambutol (98, 28, 22), isoniazid (96, 64, 41), capreomycin (33, 25, 21), para-amino salicylic acid (20, 14, 8), and bedaquiline (6, 4, 3). Delamanid and linezolid are not displayed given infrequency of use. Abbreviation: AUC_0–24, area under the concentration-time curve during a 24-hour dosing interval.

Figure 2.

A, Pharmacokinetic-pharmacodynamic pattern for the most common multidrug regimen. Clusters identified by AUC_0–24/MIC pattern at 2 weeks after treatment initiation for the regimen of pyrazinamide, levofloxacin, kanamycin, cycloserine, and ethionamide, N = 95. Pyrazinamide values represent AUC_0–24 only (pyrazinamide MIC not performed). B, Time to sputum culture conversion to negative for patients receiving the most common regimen as clustered by pharmacokinetic-pharmacodynamic pattern. Hazard ratio = 0.59 ± standard error 0.29, nominal P = .065, for difference between cluster 4 and other clusters. Abbreviations: AUC, area under the concentration-time curve during a 24-hour dosing interval; MIC, minimum inhibitory concentration.